JP2826489B2 - Combustion status detection method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a combustion state detecting method for detecting a combustion state of a burner based on an audio signal.

[0002]

2. Description of the Related Art In a case where the amount of heat generated when combustible fuel is burned is used for heating an object to be heated and utilized for industrial applications, stabilizing the combustion state of the fuel improves production efficiency. Is important above. However, when the fuel is burned, the combustion state may become unstable or fall into a misfire state due to the amount of fuel supplied to the burner, the flow rate of air, and the amount of secondary products such as steam accompanying combustion. Therefore, normally, when burning fuel, the combustion state of the burner is always detected,
Combustion conditions such as fuel supply amount, fuel flow rate, fuel pressure, supply air amount and the like are adjusted according to the combustion situation to stabilize the combustion state, and re-ignition control is performed immediately upon misfire.

Incidentally, as a method of detecting the combustion state of the burner, there is a method of detecting the combustion state based on the sound and pressure fluctuation at the time of combustion and misfire. Conventionally, as shown in FIGS. 11 and 12, for example, a microphone 31 is arranged near a flame of a burner 39 via a probe 38, and a bandpass filter 3 is obtained from an audio signal obtained by the microphone 31.
3 and an average value in a predetermined frequency region obtained by the smoothing circuit 34, and the average value is compared with a specified value from the threshold value generation circuit 36 by a comparator 35.
It detects the combustion status of the burner.

[0004]

However, in the above-described conventional combustion state detecting method, when the voice signal from the microphone 31 fluctuates due to the combustion conditions of the burner 39, the average value of these voice signals in a predetermined frequency region also fluctuates. Therefore, there is a problem that the combustion state cannot be detected with high reliability even when such an average value that tends to fluctuate and a certain specified value are compared.

Accordingly, the present invention provides a method for detecting a combustion state which can recognize a combustion state such as ignition or misfire with high reliability even when the voice signal obtained from the microphone 31 fluctuates due to the combustion conditions of the burner 39. It is intended to provide.

[0006]

In order to solve the above-mentioned object, there is provided a combustion state detecting method for detecting a combustion state of a burner based on a voice signal, and has the following features. That is, by obtaining sound signals from a plurality of systems of sound detecting means forming a pair with the burner, determining the degree of correlation between these sound signals, and comparing the degree of correlation with a specified value, the combustion state of the burner is determined. It is characterized by detecting.

According to the above arrangement, if the sound changes due to a change in the combustion condition of the burner, the sound signal obtained from the sound detecting means for detecting the sound also changes. Since the sound is formed by detecting the same voice, the amount of change has a correlation. Therefore, the correlation degree of these sound data indicates a value corresponding to the combustion state even if the sound changes due to a change in the combustion condition. As a result, even if the prescribed value used for grasping the combustion state is fixed, the degree of correlation corresponds to the combustion state, so that the combustion state can be detected with high reliability. I have.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described with reference to FIGS. For example, 10 to 40 main burners are arranged in a furnace such as a boiler furnace of a power plant.
FIG. 1 shows two of the main burners 25A and 25A ', 23 is a furnace wall, and 24 is a burner port.

In the present embodiment, a pair of microphones (in this example, two) of a probe 22 are arranged at a burner port 24 of each burner provided on the furnace wall 23. 21A ₁ · 21A ₂ is a microphone to face the probe 22 to the burner port 26 of the main burners 25A. 21A
₁ '・ 21A ₂ ' connects probe 22 to main burner 25A '
Microphone facing the burner opening 26 of the camera.

Now, the main burner 25A is ignited and the main burner 2A is fired.
Assuming that 5A 'is misfired, the microphones 21A ₁ , 21
A ₂ is to sound receiving the synthesized sound of the combustion noise SX and flow noise SY main burners 25A, will convert it into electrical audio signals. Then, when the frequency characteristic of the audio signal and SA ₁ and SA _2, respectively, the microphone 21A ₁ '· 2
In 1A ₂ ′, since the main burner 25A ′ has misfired, only the airflow sound SY is received and converted into an audio signal.

Assuming that the frequency patterns of the above audio signals are SA ₁ 'and SA ₂ ', respectively, assuming that the furnace wall 23 is made of a furnace material having a high sound absorbing property and that there is no reverberation sound, SA ₁ = SA ₂ And SA ₁ ′ = SA ₂ ′, the correlation between SA ₁ and SA ₂ (R _A1-A2 ) is (R
_A1-A2 ) = 1. Similarly, the correlation (R _{A1′-A2 ′} ) between SA ₁ ′ and SA ₂ ′ is (R _{A1′-A2 ′} ) =
It is considered to be 1. That is, the correlation of the microphone mutual frequency pattern paired is considered to be _{R A1-A2 = R A1'-} A2 '= 1. SA ₁ and SA ₁ ′, SA ₂ ′
, The degree of correlation between SA ₂ and SA ₁ ′, the degree of correlation with SA ₂ ′,
That is, it is considered that the correlation degree _{RA-A '} of the frequency patterns of the microphone on the ignition burner side and the microphone on the misfire burner side is _RA-A' = 0. Since a reverberation sound actually exists, R
_{It is considered that A'-A '} ≠ 1 (a significant value close to 1) and _RA-A' ≠ 0 (a value close to 0).

Next, the present inventors have determined that the correlation R
To confirm the relationship for _{A1'-A2 · R A1'-A1} , SA 1 and SA when ignite and misfire main burners 25A
_An experiment to determine the degree of correlation of ₂ was performed. Then, as shown in FIG. 2, and SA ₁ when the main burners 25A is ignited S
As shown in FIG. 3 and the frequency analysis result of the correlation degree of A ₂ ,
SA ₁ 'and SA ₂ ' when main burner 25A is misfired
And the frequency analysis result of the degree of correlation were obtained. During the experiment, in addition to the main burner 25A, four other main burners were also ignited at the same time, and reverberation caused by the other burners was also generated.

From the comparison between FIGS. 2 and 3, 600 to 1000
It was found that the degree of correlation in the frequency band of Hz (the range indicated by the solid arrow) was significantly different in both figures. . 4, the influence of the combustion noise of the other main burner is ignited has on the output of the sensor 21A ₁ · 21A _2, i.e., other main burner microphone output and microphone 21A ₁ · 21A
₉ shows a frequency analysis result of a degree of correlation with _two outputs. As evident from comparison of FIG. 1 and FIG. 2, in the frequency band described above, the output of the sound sensor 21A ₁ · 21A ₂ is understood that it is not affected by the combustion noise of the other main burners.

From the above considerations, (A) a pair of microphones 21A disposed on the main burner 25A
Correlation of the output of ₁ · 21A ₂ is significantly different between misfire and the time of ignition. (B) A pair of microphones 21A disposed on the main burner 25A
Correlation between the output of the arranged the sensor ₁ · 21A ₂ outputs and the other main burner, substantially equal to no, it is understood.

Therefore, the correlation between the outputs of the paired microphones arranged in the main burner in a specific frequency band is monitored, and if the correlation is a significant value, it is determined that the main burner is ignited. It was determined that if it was not a significant value, it was possible to determine that a misfire had occurred.

According to experiments by the present inventors, it was sufficiently practical to set 0.4 as the significant value. Further, when the main burners 25A is complete misfire (i.e., even flow noise since there is no feed to the furnace of the combustion air, there is no state combustion noise), the sound microphone 21A ₁ · 21A ₂ is due to other main burner And the degree of correlation (referred to as R _{A0 -A 0} ) increases as shown in FIG. 5, but since this “complete misfire” is usually performed artificially, No distinction is required.

[0017] Accordingly, desirably the electric audio signal outputted by the microphone 21A ₁ · 21A ₂ are determined by calculating the correlation of a specific frequency component, and a specified value the correlation (e.g., above 0.4) By comparing
It became clear that the combustion status such as ignition / misfire can be automatically detected.

Next, the configuration of a combustion state detecting device for detecting a combustion state by executing the above procedure will be described.

As shown in FIG. 6, the combustion state detecting device has two systems of sound detecting units 1.1. Each of the sound detectors 1 is arranged in a left and right pair with respect to a flame generator (not shown), and includes a microphone probe 2 including a microphone and a probe for detecting a sound signal, and an amplifier 3 for amplifying the sound signal detected by the microphone probe 2. And a high-pass filter 4 that passes a high-frequency component (frequency region equal to or higher than the frequency fh) in the audio signal amplified by the amplifier 3, and a low-frequency component (frequency region equal to or lower than the frequency fl) in the audio signal that has passed through the high-pass filter 4. ) Is passed through the low-pass filter 5.

A trimmer 6 is connected to each of the amplifier 3 and the filters 4 and 5. By operating each of the trimmers 6, the amplification factor of the amplifier 3 and each of the filters 4 and 5 are controlled.
5, the frequency fh · fl is finely adjusted. It is preferable that a switch such as a dip switch that can switch the amplification factor and the frequency fh · fl stepwise be connected to the amplifier 3 and the filters 4 and 5.

The two sound detectors 1 and 1 are connected to an arithmetic processing board 7 for judging the combustion state based on the sound signal. The arithmetic processing board 7 includes an A / D converter 8 which converts an audio signal into audio data of a digital value and takes in the data, and a CPU (Cent
ral Prosessing Unit) unit 9 and a DSP (Dig) that performs arithmetic processing such as FFT conversion during data processing in cooperation with the CPU unit 9.
Ital Signal Prosessor (PI) unit 10, a PI (Parallel Input) unit 11 to which external data is input in parallel, and a PO (Parallel Outpu) that outputs evaluation values after data processing in parallel.
t) unit 12 and a D / A conversion unit 13 that converts an evaluation value into an analog value and outputs the analog value.

The A / D converter 8 is connected to the voice detectors 1 and 1.
1, an audio signal is input. The A / D converter 8 converts the audio signal into audio data and then outputs the audio data to the CPU 9. Also,
A processing changeover switch 14 such as a dip switch is connected to the PI unit 11, and the processing changeover switch 14 inputs a switching data signal designating the processing content of the audio data to the CPU unit 9 via the PI unit 11. It has become.

CPU to which audio data and switching data signals are input from A / D converter 8 and PI unit 11
The unit 9 accesses the DSP unit 10 as needed to perform arithmetic processing, and performs data processing on the audio data in accordance with the processing specified by the switching data signal.

That is, the CPU section 9 executes two types of data processing based on the switching data signal. In the first data processing, FFT (Fast Fourier Transform) conversion is performed on each voice data obtained from the two voice detection units 1 and 1 to obtain FFT spectrum data, and then the degree of correlation between these FFT spectrum data is obtained. Is obtained, and an integrated value of coherence values in a predetermined frequency region is obtained as an evaluation value.

In the second data processing, each of the voice data obtained from the two systems of voice detectors 1 and 1 is subjected to FFT to obtain FFT spectrum data, and then the degree of correlation between these FFT spectrum data is obtained. Are obtained, and the integrated values of the coherence values in two different frequency regions are obtained, and the ratio of these integrated values is obtained as the evaluation value.

The CPU section 9 is connected to the D / A conversion section 13 and the PO section 12, and outputs an evaluation value to the D / A conversion section 13 and the PO section 12, respectively. . The D / A conversion unit 13 is connected to two analog output systems 19, 19, and converts the evaluation value into an analog signal to each output system 19 and outputs the analog signal. Each analog output system 19 includes an amplifier 15 for amplifying an output signal and a trimmer 1 for finely adjusting the amplification factor of the amplifier 15.
6 and a display 17 for displaying the voltage value and the like of the output signal output from the amplifier 15. The output state is output to the outside while being checked by the display 17. On the other hand, the PO unit 12 has four digital output systems 2.
, And outputs the evaluation value to each output system 20 as a parallel digital signal. Each digital output system 20 is provided with the lamp 18 so that the output state is output to the outside while the output state is confirmed by the lamp 18.

The analog output system 19 and the digital output system 20 are connected to a discriminating device or the like (not shown), and the discriminating device compares the evaluation value (integrated value of coherence value, ratio of integrated value) with a specified value. By comparing, the combustion state is grasped.

The operation of the combustion state detecting device in the above configuration will be described. When the combustion is started, the sound accompanying the combustion is detected by the microphone probes 2 and 2 and taken into the sound detection unit 1 1 as an electric sound signal. After being amplified by the amplifier 3, the audio signal of each audio detector 1 is converted into a component only in a predetermined frequency region by the high-pass filter 4 and the low-pass filter 5. Then, the data is output to the A / D converter 8 of the arithmetic processing board 7, converted into digital-valued audio data by the A / D converter 8, and then output to the CPU 9.

When the CPU 9 captures both voice data from the voice detectors 1 and 1, the CPU 9 cooperates with the DSP 10 according to the processing content specified by the switching data signal from the processing switch 14. While processing data.

Specifically, for example, when the switching data signal is the first
As shown in FIG. 7, for example, as shown in FIG. 7, each voice data at the time of main burner combustion, main burner misfire, complete fire extinguishing, etc. is subjected to FFT conversion to obtain FFT spectrum data. Become. Thereafter, as shown in FIG. 8 to FIG.
A healence value indicating the degree of correlation of the spectrum data is calculated.

That is, the method of calculating the coherence value will be described in detail. It is assumed that the signal Y _k is given by the equation (1) for the signal X _k .

[0032]

(Equation 1)

N _k is a noise signal component having no correlation with X _k . Here, the cross power spectrum X _k Y _k ^*
Is represented by the equation (2). Note that X _k and Y _k are complex numbers, and a numerical value with “*” is a conjugate complex number.

[0034]

(Equation 2)

The average value of M times in equation (2) is represented by equation (3), where () m belongs to the m-th sample sequence.

[0036]

(Equation 3)

Then, the second term on the right side of the equation (3) becomes "0" if M → ∞. Therefore,
If M is set to a sufficiently large value, equation (4) will be obtained. It should be noted that “─” attached above the numerical value indicates an average value.

[0038]

(Equation 4)

If V _k = H _k X _k , the coherence value (r ² (k)) is given by equation (5) according to the definition equation.

[0040]

(Equation 5)

Here, X _k = X _r (k) + jX _i (k), Y
_k = a _{Y r (k) + jY i} (k). Furthermore, X _r a simplified X _r (k), likewise Y _r (k) the Y _r, X _i and (k) X
_{When i} and Y _i (k) are represented by Y _i , the numerator portion is represented by Expression (6), and the denominator portion is represented by Expression (7).

[0042]

(Equation 6)

[0043]

(Equation 7)

Further, by further averaging the expressions (6) and (7), the coherence value (r ² (k)) becomes
It will be defined by equation (8).

[0045]

(Equation 8)

Thus, the F obtained from the audio data
X based on FT spectrum data _r, X_i, Y_r, Y
_iAre obtained, these values are substituted into the above equation (8).
The FFT spectrum data
The reference value will be calculated.

When the coherence value is calculated as described above, a predetermined frequency region (f1 to f1) is calculated from the following equation (9).
The integrated value (evaluation value R) of the coherence value of f2) is obtained. Note that f1 and f2 indicate the lower limit frequency and the upper limit frequency of the region including the flame combustion noise, for example, in the range of 800 Hz to 1 kHz in FIGS. 8 and 9.

[0048]

(Equation 9)

Thereafter, the above integral value is used as an evaluation value R as C
It is output from the PU unit 9 to the analog output system 19 and the digital output system 20 via the D / A conversion unit 13 and the PO unit 12, and the D / A conversion unit 13 converts the evaluation value R into an analog signal. After that, the signal is output to each analog output system 19. On the other hand, the PO unit 12 outputs the evaluation value R to each digital output system 20 in a parallel format. Then, the evaluation value R is output to the discriminating device (not shown) via the analog output system 19 and the digital output system 20, and is compared with a specified value in the discriminating device, so that it is used for grasping the combustion state.

On the other hand, if the switching data signal specifies the second data processing, the coherence value is obtained as in the case of the first data processing, and then the first frequency domain is obtained from the following equation (10). The ratio between the integral value of the coherence value of (fh1 and fh2) and the integral value of the coherence value of the second frequency region (fl1 and fl2) is obtained as the evaluation value R. Note that fh1 and fh2 are, for example, as shown in FIG.
9 shows a lower limit frequency and an upper limit frequency of a high frequency region including the flame combustion noise, respectively, as in the range of 800 Hz to 1 kHz in FIG.
2 is, for example, in the range of 400 Hz to 600 Hz,
FIG. 4 shows a lower limit frequency and an upper limit frequency of a low frequency region that does not include a flame combustion sound.

[0051]

(Equation 10)

Thereafter, the evaluation value R is output from the CPU unit 9 to the D / A conversion unit 13 and the PO unit 12, and the D / A conversion unit 13 converts the evaluation value R into an analog signal. After that, the signal is output to each analog output system 19. On the other hand, the PO unit 12 outputs the evaluation value R to each digital output system 20 in a parallel format. The evaluation value R is calculated based on the analog output system 19 and the digital output system 2.
The signal is output to a discriminating device (not shown) through 0, and is compared with a specified value in the discriminating device to be used for grasping the combustion state.

As described above, according to the combustion state detecting method of the present embodiment, audio signals are obtained from a plurality of systems of audio detecting means forming a pair with the burner, the correlation of these audio signals is obtained, and the correlation is obtained. Is compared with a specified value to detect the combustion state of the burner. More specifically, audio signals are obtained from a plurality of systems of audio detection means (such as the microphone probe 2) paired with the burner, coherence values are obtained from FFT spectrum data of these audio signals, and the coherence values are determined based on the coherence values. The combustion state of the burner is detected by comparing the evaluation value obtained as described above with a specified value.

As a result, when the sound changes due to a change in the combustion conditions, the sound signal obtained from the sound sensor for detecting the sound also changes. However, these sound signals are detected by detecting the same sound. Because it was formed,
The amount of change has a correlation. Therefore, the degree of correlation of these audio signals in a specific frequency band indicates a value corresponding to the combustion state even if the audio changes due to a change in combustion conditions. As a result, even if the specified value used for grasping the combustion state is fixed, the degree of correlation corresponds to the combustion state, so that the combustion state can be recognized with high reliability. I have.

Further, the combustion state detecting method of this embodiment is as follows.
The combustion state is monitored using the ratio of the degree of correlation in different frequency regions as an evaluation value. This makes it possible to cancel the variation with respect to the uncorrelated noise by the denominator and the numerator by using the correlation degree ratio as the evaluation value, so that the white noise ( Even if a large amount of electrical noise components are mixed in with the combustion noise, the combustion state can be recognized with high reliability.

[0056]

As described above, according to the present invention, audio signals are acquired from a plurality of systems of audio detection means forming a pair with a burner, the degree of correlation of these audio signals is obtained, and the degree of correlation is defined as a specified value. The configuration is such that the combustion state of the burner is detected by comparison.

Thus, even if the sound changes due to a change in the combustion condition, the correlation value of the sound data indicates a value corresponding to the combustion state, so that the prescribed value used for grasping the combustion state is fixed. However, there is an effect that the combustion state can be recognized with high reliability.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention.

FIG. 2 is a diagram showing a correlation-frequency relationship between a sensor and an output during burner combustion.

FIG. 3 is a diagram showing a correlation-frequency relationship between a sensor and an output when a burner misfires.

FIG. 4 is a diagram showing a relationship between a burner sensor output and a sensor output correlation degree-frequency of another burner.

FIG. 5 is a diagram showing a correlation-frequency relationship between sensors and outputs when a burner is extinguished.

FIG. 6 is a block diagram of a combustion state detecting device.

FIG. 7 is a graph showing speech FFT spectrum data.

FIG. 8 is a graph showing a coherence value during main burner combustion.

FIG. 9 is a graph showing a coherence value when a main burner misfires.

FIG. 10 is a graph showing a coherence value at the time of complete fire extinguishing.

FIG. 11 is a block diagram showing a conventional acoustic ignition detection device.

FIG. 12 is a layout diagram of main parts in the acoustic ignition detection device.

[Explanation of symbols]

Reference Signs List 1 voice detection unit 2 microphone probe 3 amplifier 4 high-pass filter 5 low-pass filter 6 trimmer 7 arithmetic processing board 8 A / D conversion unit 9 CPU unit 10 DSP unit 11 PI unit 12 PO unit 13 D / A conversion unit 14 processing changeover switch 15 Amplifier 16 Trimmer 17 Display 18 Lamp 19 Analog output system 20 Digital output system 25A Main burner 21A ₁ sound sensor 21A ₂ sound sensor

──────────────────────────────────────────────────の Continued on the front page (72) Inventor Satoru Taniguchi 1-5-5 Takatsukadai, Nishi-ku, Kobe-shi, Hyogo Kobe Steel, Ltd. Kobe Research Institute (72) Inventor Tatsuo Kono Nishi-ku, Kobe-shi, Hyogo 1-5-5 Takatsukadai Kobe Steel, Ltd.Kobe Research Institute (72) Inventor Takehiko Nakata 2-3-1, Shinhama, Araimachi, Takasago City, Hyogo Prefecture Kobe Steel, Ltd. Takasago Works (72) Invention Person Adachi Adult 2-3-1 Shinhama, Arai-machi, Takasago City, Hyogo Prefecture Inside Kobe Steel, Ltd. Takasago Works (58) Field surveyed (Int. Cl. ⁶ , DB name) F23N 5/24 G01H 17/00 F27D 21 / 00 F27D 7/06

Claims (1)

(57) [Claims] 1. A combustion state detection method for detecting a combustion state of a burner based on a sound signal, wherein sound signals are obtained from a plurality of systems of sound detection means forming a pair with the burner, and a correlation between the sound signals is obtained. Ask for degree,
A combustion state detection method, wherein the combustion state of the burner is detected by comparing the degree of correlation with a specified value.